Part 1 - NN as Regressor

All the parameter values are in float except the Signal strength

Parameter 3 ranges between 0 and 1.

Maximum value of Parameter 5 is 0.6

Parameter 8 has a very low range between 0.9 and 1.004 Standard deviation is lowest for Parameter 8, it is 0.001887

1.Parameter 6 and Parameter 7 are highly correlated with each other and visce versa and they have almost 0 correlation with other Parameters

2.Parameter 1 is positively correlated to Parameter 3 and Parameter 8 and negatively correlated to Parameter 2 and Parameter 9.

3.Parameter 4 is has very low correlation with other Parameters.

Since high correlation coefficient value lies between ± 0.50 and ± 1 Parameter 1 is highly correlated with Parameter 3 and Parameter 8, Parameter 9. Parameter 6 and 7 are highly correlated. But since, the correlation is not too high near 0.8 or above not dropping the features.

Parameter 4 has the highest number of outliers which is 155.

Part 2 - Autonomous Vehicles Project

Read Data

Data Analysis and Preparation

Visualize and Standardize the data

Need to reshape the X_train and X_test so that the same can be fed for model building. We need to feed a 2D tensor into the model and currently we have a 3D tensor.

Encoding the target variables. We need to one hot encode the labels for the model to understand the labels better. We will be using categorical cross entropy as our loss function and for this purpose we need our labels to be in one hot encoded format.

Model building

Conclusion and Interpretation

There are around 10 classes in the dataset which represent digits from 0-9.

We tried training a Neural Network with dense hidden layers of different number of units and are able to achieve a final test accuracy of 80.77 %.

Also we notice that after a certain point the model begins to overfit on our dataset as is clear from the plots above where the validation loss begins to increase after certain point and validation accuracy begins to decrease.

Thus, with this amount of accuracy we are able to distinguish between the different digits in this dataset.